4 ms·
> At one point after days of training, my agent actually had a 90% LOSE rate against an opponent that played totally randomly. Nice! You should have just added
by philipov 10y ago
> At one point after days of training, my agent actually had a 90% LOSE rate against an opponent that played totally randomly.
Nice! You should have just added a bit at the end to invert whatever answer it got, and you would have had a winner.
But more seriously, I think that we will become more clever with designing genetic algorithms to evolve the neural networks as part of the training process rather than trying to build our own from scratch every time. I vaguely recall there is some research being done on that front already.
- deepnotderp 10y agoNeural architecture search with reinforcement learning. We've used an actor critic method internally with good results as well.
- mrjoeblack 10y agoThe fact that a random opponent performs better means that simply inverting the output of a bad strategy (assuming that is even possible, in cases where the output is more complex than binary it should not be) would just give you another bad strategy.
- philipov 10y ago> ...would just give you another bad strategy. That's the joke, but to be fair, inversion is not a binary concept. Negation is binary, but inversion is more general and has the 2D interpretation of reflecting something across X=Y.